Analyses

Confirmatory factor analysis

Test a factor model you specified in advance.

Edit on GitHub

When to use it

Use confirmatory factor analysis when you already claim that certain items measure a named factor. The model is a sentence such as F1 =~ q1 + q2 + q3 + q4. Fit indices say whether that claim is tolerable. To hunt for a structure, use exploratory factor analysis.

Assumptions

Indicators are numeric. Rows with a missing indicator are dropped. The model has to be identified: a factor with a single indicator, or too few indicators for the number of free loadings, will not fit. The first loading on a factor is fixed at 1 so the factor has a scale. Tensr does not offer an estimator menu.

Running it in Tensr

Multivariate → CFA. In chat: “CFA with F1 measured by q1, q2, q3, and q4.”
List the indicator columns, at least two.
Write the model. Example: F1 =~ q1 + q2 + q3 + q4.

Options

Prop

Type

Reading the output

A one-factor model, Focus, measured by item1 to item4. Those four items share a moderate factor.

IndicatorFactorEstimateStd. Errp-value
item1Focus1——
item2Focus1.404380.425655p < .001
item3Focus1.379480.418823p < .001
item4Focus0.9435590.319745p = .003

Confirmatory factor analysis with semopy fit indices. Confirmatory factor analysis with semopy fit indices. Metrics: CFI = 0.999; RMSEA = 0.014; SRMR = 0.04.

Reporting (APA 7)

Confirmatory factor analysis with semopy fit indices. Report the estimate in the table. This procedure is not summarised by one p-value.

Coming from SPSS

The product menu is Multivariate → CFA. SPSS itself fits this kind of model in Amos, not in the Analyze menu. The syntax is the same shape as lavaan: F1 =~ q1 + q2 + q3 + q4.